Classification of Online Judge Programmers based on Rule Extraction from Self Organizing Feature Map

Computer programming is one of the most important and vital skill in the current generation. In order to encourage and enable programmers to practice and sharpen their skills, there exist many online judge programming platforms. Estimation of these programmers' strength and progress has been an important research topic in educational data mining in order to provide adaptive educational contents and early prediction of `at risk' learner. In this paper, we trained a Kohonen Self organizing feature map (KSOFM) neural network on programmers' performance log data of Aizu Online Judge (AOJ) database. Propositional rules and knowledge was extracted from the U-matrix diagram of the trained network which partitioned AOJ programmers into three distinct clusters ie. `expert', `intermediate' and `at risk'. The proportional rules performed classification with an accuracy of 94% on a testing set. For validation and comparison, three more predicting models were trained on the same dataset. Among them, feedforward multilayer neural network and decision tree have scored accuracy of 97% and 96% respectively. In contrast, the precision score for support vector machine was about 88%, but it scored the highest recall score of 99% in terms of identifying `at risk' students.